Logdecay: Difference between revisions

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search
imported>Scott
imported>Donal
 
(2 intermediate revisions by 2 users not shown)
Line 1: Line 1:
===Purpose===
===Purpose===


Variance scales a matrix using the log decay of the variable axis.
Scales a matrix using the inverse log decay of the variable axis.


===Synopsis===
===Synopsis===
Line 11: Line 11:
Inputs are data to be scaled (x), and the decay rate (tau). Outputs are the variance scaled matrix (sx) and the log decay based variance scaling parameters (logscl).
Inputs are data to be scaled (x), and the decay rate (tau). Outputs are the variance scaled matrix (sx) and the log decay based variance scaling parameters (logscl).


For an m x n matrix 'x' the variance scaling used for variable 'i' is exp(-(i-1)/((n-1)\*tau)). This gives a scaling of 1 on the first variable (i.e. no scaling), and a scaling of 1/exp(-1/tau) on the last variable. The following table gives example values of tau and the scaling on the last variable:
For an m x n matrix 'x' the variance scaling used for variable 'i' is exp(-(i-1)/((n-1)\*tau)). This gives a scaling of 1 on the first variable (i.e. no scaling), and a scaling of 1/exp(-1/tau) on the last variable. The following table gives example values of tau and the scaling which is applied to the last column of x (the last variable):


{| class="wikitable"
{| class="wikitable"
Line 29: Line 29:
===See Also===
===See Also===


[[autoscale]], [[scale]]
[[auto]], [[scale]]

Latest revision as of 10:50, 7 June 2016

Purpose

Scales a matrix using the inverse log decay of the variable axis.

Synopsis

[sx,logscl] = logdecay(x,tau)

Description

Inputs are data to be scaled (x), and the decay rate (tau). Outputs are the variance scaled matrix (sx) and the log decay based variance scaling parameters (logscl).

For an m x n matrix 'x' the variance scaling used for variable 'i' is exp(-(i-1)/((n-1)\*tau)). This gives a scaling of 1 on the first variable (i.e. no scaling), and a scaling of 1/exp(-1/tau) on the last variable. The following table gives example values of tau and the scaling which is applied to the last column of x (the last variable):

tau scaling
1 2.7183
1/2 7.3891
1/3 20.0855
1/4 54.5982
1/5 148.4132

See Also

auto, scale